VQ-VAE-2

VQ-VAE-2 is a hierarchical vector quantized variational autoencoder designed to improve generative modeling of images by combining multi-scale latent representations. It extends its predecessor by introducing a two-level latent hierarchy, enabling high-quality image generation and efficient compression.

Read More →

Perplexity

Perplexity is a measurement used primarily in the fields of information theory and natural language processing to quantify how well a probability model predicts a sample. It reflects the uncertainty or unpredictability of a probability distribution, often applied in evaluating language models.

Read More →